Pattern-based Wordiness Reduction System for Thai Texts

In this paper, we propose a method and web application to reduce wordiness in Thai by adapting pattern-based analysis with wordiness initial corpus. The small amount of Thai text is gathered to create the initial corpus for grammatically analyzing the patterns of Thai wordiness. 18 wordiness have been decided to intensely analyze the co-occurrence, patterns contexts and construction rules. And 99 wordiness patterns are discovered for the experiments. The experiment has been done with 100 documents from Thai National Corpus by 15 students. For the user satisfaction which is analyzed from the questionnaire, users are highly satisfied with the use of the proposed system. And the wordiness reduction result of the experiment shows the effectiveness of the wordiness identification and reduction which the accuracy is 86.51%. The 13.49% of errors come from word segmentation and unknown words. Wordiness automatically corpus will be developed for further experiment in the future.